JPH07218565A - Detection method for component with jitter frequency - Google Patents

Abstract

PURPOSE:To judge whether the peak waveform of spectrum with jitter frequency is attributed to aliasing or not. CONSTITUTION:A time signal x(t) is computed by interpolator computation of the interval of time series data P1, P2..., which does not necessarily have equal interval, sampling is carried out at slightly different two sampling frequencies fs1, fs2=fs1 + or -DELTAfs, DFT(discrete Fourier transformation) is carried out to compute spectrum XI (f=0-fs1/2) with Jitter discrete frequency and spectrum X2 (f=0-fs2/2) and also compute the frequency f11<f12<...<flm and f21<f22<...<f2m' at which respective peaks can have respectively prescribed values, and the peaks at f1i f2i are presumed to be true peaks and the peaks at f2i f1i+ or -DELTAfs are presumed to be attributed to aliasing. As the DFT, an FFT (fast Fourier transformation) can be used. Continuous data, sampling data, or migration average value can be used as the time series data P1, P2,.... It is preferable that the spectrum is displayed while the peak waveforms attributed to aliasing are eliminated or that the spectrum is displayed while the parts attributed to aliasing are made obvious.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention finds a continuous time signal x (t) by interpolating time-series data of a certain physical quantity, which is not necessarily at regular intervals, and samples x (t) at a period ts. To obtain an evenly spaced discrete time signal x (nts) and perform a discrete Fourier transform (DFT) operation on the equally spaced discrete time signal x (nts) to obtain a discrete frequency spectrum X.
The present invention relates to a method for detecting a jitter frequency component of a physical quantity so as to obtain (nΔf). As the DFT, a fast Fourier transform (FFT) that shortens the calculation time is often used.

[0002]

2. Description of the Related Art As shown in FIG. 5, a jitter measuring apparatus 1 used in a conventional jitter detecting method of this type has an input terminal INa.
Alternatively, the signal input to INb is subjected to level adjustment, waveform shaping, etc. in the measurement signal generation circuit 2 as necessary, and for example, the signal v (t) of FIG. 6A is input to the period measurement circuit 3a or 3b, respectively. The period data T ₁ , T _2, ... And data acquisition times t ₁ , t _2, ... Of each signal are continuously measured, and the measured data are written in the memory 4a or 4b, respectively. In the measurement amount calculation unit 5, corresponding data P ₁ , P ₂ , and P ₂ according to the type of physical quantity (cycle, frequency, pulse width, duty ratio, time interval, phase, etc.) specified from the cycle data Ti of the memory 4a or 4b. P ₃ (FIG. 6B) is calculated as needed. Then, the obtained non-equidistant data P ₁ , P ₂ , P _3, ... Are input to the frequency component analysis unit 6.

The frequency component analysis unit 6 performs an interpolating operation between the data of the input non-equidistant data string {Pi} and the data, that is, by connecting with an interpolation curve to obtain a continuous interpolated waveform of the physical quantity, that is, The time signal x (t) (Fig. 6B) is determined. In order to calculate the discrete frequency spectrum of the jitter by the DFT, data at equal intervals are required. Therefore, as shown in FIG. 6C, the sampling frequency f
s (sampling period ts = 1 / fs) is sampled and N pieces of data Q ₀ , Q ₁ ,
Q ₂ , ... Q _N-1 are extracted. That is, Qn = x (nts); n = 0, 1, 2, ... N-1 (1) The sampling frequency fs is more than twice the upper limit of the band in which the frequency spectrum of the jitter is to be measured according to the sampling theorem. Is selected as the frequency. In the DFT theory, the N pieces of the extracted time series data Q _{0 to} Q _N−1 are regarded as those that are repeatedly generated, and the calculation is performed. As described above, the sampling data extraction range and its time length Nts are set in consideration of the change characteristics of the time signal x (t).

The equally-spaced data Q extracted in this way
_The DFT operation is performed using _{0 to} Q _N−1, and the discrete frequency spectrum X (nΔf) as shown in FIG. 6D; n = 0,
1, 2, ... N-1 are obtained. Δf is Δf = fs / N (2) If the spectral data is set to Dn, Dn = X (nΔf); n = 0, 1, ..., N-1 (3) The discrete frequency spectrum (amplitude) X is It has a characteristic that it is bilaterally symmetrical with respect to the frequency f, that is, f = 0 (having symmetry), and the spectrum X has a frequency interval fs.
The spectrum X (f = 0
.About.fs), symmetrical waveforms S and S ^* are obtained around fs / 2 (referred to as Nyquist frequency), and the spectrum up to the vicinity of Nyquist frequency fs / 2 is displayed on the display unit 7.

The number N of data to be sampled from x (t)
Indicates that the amount of DFT operation increases in accordance with N, and that the frequency spectrum X is 0 / fs / 2 with N / at intervals of Δf.
It must be selected in consideration of the fact that it is displayed as two data.

[0006]

If the frequency spectrum X (f = 0 to fs / 2) obtained by the DFT operation has a spectrum component with a Nyquist frequency of fs / 2 or more, this is the point of fs / 2. It will be folded back at, and it will be a characteristic that is superimposed on the existing spectrum. The folding and overlapping of the spectra of fs / 2 or more is called aliasing.

Now, the physical quantity represented by the time signal x (t) is taken as a frequency, and this frequency waveform x (t) is x (t) = fc + ΔF cos2πPt (4) fc = FM frequency center frequency (carrier) ΔF = frequency Consider the case where the maximum deviation P = modulation signal frequency.

(B) fc = 10, ΔF = 2, P = 3
(KHz) (FIG. 2A) (b) fc = 10, ΔF = 2, P = 7 (kHz)
Consider the spectrum for (FIG. 3A). As for obtaining the spectrum (jitter component) of f = 0 to 5 kHz, fs ₁ = 1
Set to 0 kHz and sample value x (nts ₁ ); (n =
0, 1, 2, ..., N-1; ts ₁ = 1 / fs ₁ ) is extracted,
DFT operation is performed to obtain spectrum X ₁ (nΔf ₁ ) (Δf
₁ = fs ₁ / N) is calculated.

In the case of (a), X ₁ (nΔf ₁ ) is as shown in FIG.
As shown in B, a waveform S ₁ having a peak at f1 ₁ = P = 3 kHz and a waveform S centering on the Nyquist frequency fs _1/2
1 and the symmetrical waveform S ₁ ^* is obtained. For (b), the Nyquist frequency fs _1/2 = 5kHz or more jitter component P
= 7 kHz is an example. The spectrum X ₁ (nΔf ₁ ) obtained by the DFT calculation is as shown in FIG. 3B.
The waveform centered around the jitter frequency P = 7 kHz is fs ₁ /
Folded in terms of _{2, fs 1 / 2- (P} -fs 1/2) =
A peak waveform S _1e due to aliasing having a peak at a point of fs ₁ −P = 10 −7 = 3 kHz appears, and fs ₁ /
A peak waveform S _1e ^* having a peak at 7 kHz appears, which is symmetrical with respect to the peak waveform S _1e around 2. However, the display unit 7 displays the range of f = 0 to fs _1/2 for both (a) and (b).

The peak waveforms S ₁ and S _1e of the discrete frequency spectrum shown in FIGS. 2B and 3B have almost the same characteristics. The characteristics of equation (4) and (a) and (b) relating to the above-mentioned frequency waveform x (t) cannot be accurately known in advance before actually obtaining the frequency spectrum, so see FIG. 2B or FIG. 3B. There is a problem that it is impossible to determine whether it is the peak waveform of aliasing or not by just doing it.

An object of the present invention is to make it possible to determine whether or not the peak waveform of the discrete frequency spectrum X of jitter is due to an aliasing peak.

[0012]

[Means for Solving the Problems]

(1) In the jitter frequency component detecting method according to the first aspect, continuous time signals x (are connected by interpolating adjacent data with time-sequential non-equidistant or equidistant digital data relating to an arbitrary physical quantity. t) (t is time) is obtained, and slightly different first and second sampling frequencies fs ₁ , fs ₂ = fs ₁ + Δfs are set for the time signal x (t). The first and second discrete time signals (equal interval discrete data) are sampled from the time signal x (t) at the first and second sampling periods ts ₁ = 1 / fs ₁ and ts ₂ = 1 / fs _2. Row)
x (nts ₁ ), x (nts ₂ ) (n = 0, 1, 2, ..., N
-1) is extracted. Their first and second discrete time signals x
_{(Nts 1), x (nts} 2) by performing a discrete Fourier transform (DFT) operation on the first and second discrete frequency spectrum _{X 1 (f = 0~fs 1/} 2), X 2 (f = 0 ~ Fs ₂
/ 2) is calculated.

The first and second frequency spectra
X₁, X₂Frequency f at which the amplitude of the signal reaches a peak above a predetermined value
1₁<F1₂<... <f1m and f2₁<F2₂<... <f2m it
Obtain each of the first and second frequency spectra X₁, X₂
The peak frequencies f1i and f2i (i = 1 to m) are compared,
The peak at which f1i ≈ f2i (i = 1, 2, ..., M) is true
The peak is determined and f2i ≈ f1i ± Δfs (i = 1, 2,
,, m) is determined to be the peak due to aliasing.
Set. (2) In the invention of claim 2, the time system relating to an arbitrary physical quantity
For non-uniform or evenly-spaced digital data {Pi} in columns
In contrast, slightly different first and second sampling frequencies
fs₁, Fs₂= Fs₁Set ± Δfs and set the time series
From the digital data {Pi}, the first and second sample
Cycle ts₁= 1 / fs₁, Ts₂= 1 / fs₂Each service
Extract the data existing near the sampling time
Create 1st and 2nd time series data {Ui}, {Vi}
It The first and second time series data {Ui}, {V
i} is interpolated, and the time interval is the first and second sample
Long cycle ts₁, Ts₂The first and second equal to
Equidistant data string (discrete time signal) x ₁(nts₁), X
₂(Nts₂) (N = 0, 1, 2, ... N-1),
After that, the same processing as the above item (1) is performed. (3) In the invention of claim 3, the above (1) or (2)
In the method of detecting the jitter frequency component shown in
D) Transform (DFT) is called Fast Fourier Transform (FFT)
It

(4) In the invention of claim 4, in the method of detecting a jitter frequency component according to any one of (1) to (3), the digital data is sampling data. (5) In the invention of claim 5, in the method of detecting a jitter frequency component according to any one of (1) to (3), the digital data is a moving average value of the physical quantity.

(6) According to the invention of claim 6, in the method of detecting a jitter frequency component according to any one of (1) to (5), the peak characteristic judged as the aliasing is deleted and the frequency is eliminated. The spectrum X ₁ or X ₂ is displayed as an image. (7) In the invention of claim 7, in the method of detecting a jitter frequency component according to any one of (1) to (5), which part of the spectrum is aliased together with the frequency spectrum X ₁ or X _2. This is an image display of whether the peak waveform is due to.

[0016]

Embodiments The cases (a) and (b) described in "Problems to be solved by the invention" will be described as an example. Frequency f = 0
Setting the fs ₁ = 10 kHz as seeking ~fs _1/2 of the frequency spectrum (jitter), but further fs ₁ and slightly different second using the sampling frequency fs ₂ in the present invention. Now let it be fs ₂ = 11 kHz. Therefore Δ
fs = fs ₂ -fs ₁ = is 1kHz.

Case (a) Discrete frequency spectrum X ₁ (nΔf ₁ ) (n = 0, 1, 2, ..., When sampling frequency fs ₁ = 10 kHz)
N-1) is as already shown in FIG. 2B. Frequency fs ₂
N discrete data x (nts ₂ ) (n
= 1 to N-1) by performing a DFT operation on a discrete frequency spectrum X ₂ (nΔf _2); (but Δf ₂ = fs ₂
/ N) is obtained as shown in FIG. 2C. f2 ₁ = P = 3k
The peak waveform S _{2 of the} jitter whose peak is Hz and fs ₂ /
2 is a waveform S ₂ and symmetrical waveform S ₂ ^* exists around the. This is natural because the frequency signal x (t) is expressed by the equation (4). In the Nyquist frequency fs _2/2 or less of the display range, the peak frequencies f1 _1, f2 ₁ peak waveform S ₁ and S ₂ are _{f1 1 = f2 1 = P ...} (5), be changed sampling frequency It does not change.

In the case of (b) The spectrum X ₂ (nΔf ₂ ) when fs ₁ = 10 kHz
(N = 0, 1, 2, ..., N-1) have already been shown in FIG. 3B. N pieces of data x (nts ₂ ) (n = 0 to N−1) sampled at the sampling frequency fs ₂ = 11 kHz
The spectrum X ₂ (nΔf
₂ ); (but Δf ₂ = fs ₂ / N) is as shown in FIG. 3C. The jitter component of P = 7 kHz is the Nyquist frequency fs ₂
Folded at /2=5.5 kHz, f2 ₁ = fs ₂ -P =
The peak waveform S _2e due to aliasing appears at 4 kHz,
The fs _2/2 around the peak waveform S _2e and symmetrical peak waveform S _2e ^* appears in a point P = 7 kHz.

For f1 ₁ = fs ₁ -P = 3 kHz, f2 ₁
= Fs ₂ −P = 4 kHz and Δfs = 1 kHz, it can be seen that f2 ₁ = f1 ₁ + Δfs (6) is satisfied. As described above, the peak frequency f1 ₁ due to the aliasing in FIG. 3B can be obtained by shifting the sampling frequency fs ₁ by ± Δfs and the folding point fs ₁ /
Since 2 is shifted by ± Δfs / 2, f2 ₁ = f1 ₁ ± Δfs
It is possible to determine that the peak is due to aliasing by shifting to.

[0020] In general, the Nyquist frequency fs _1/2, fs 2 /
In the case of m (integer of 2 or more) pairs of peak frequencies f1 ₁ <f1 ₂ <... <f1m; f2 ₁ <f2 ₂ <... <f2m ... (7) , Peak frequencies f1i and f2i (i = 1, 2,
, M) and f1i ≈ f2i (i = 1, 2, ..., M) (8) is determined to be a true peak, and f2i ≈ f1i ± Δfs (i = 1, 2, , M) (9) can be determined as a peak due to aliasing.

FIG. 1 is obtained by summarizing the method of detecting the jitter frequency component of the physical quantity of the present invention described above. (Modification; Claim 2) A modification will be described with reference to FIG. Two sampling frequencies fs ₁ and fs ₂ = fs ₁ ± Δ that are slightly different from the time-series digital data {Pi}
Set fs (step S ₀ ). Next, the first time series data {Ui} in the vicinity of each sampling time of the sampling period ts ₁ = 1 / fs ₁ is extracted from the time series data {Pi} (step S ₁ ). Similarly, time series data {Pi}
Then, the second time series data {Vi} in the vicinity of each sampling time of the sampling cycle ts ₂ = 1 / fs ₂ is extracted (step S ₂ ).

First and second time series data {Ui}, {V
i} respectively by interpolating, respectively equal discrete time signal time interval is the sampling period ts _1, ts ₂ (equidistant discrete data _{sequence) x 1 (nts 1),} x 2 (nts 2) (n = 0,
1, 2, ..., N-1) are extracted (step S ₃ ,
S ₄ ). Omitted because the processing of the subsequent steps S ₅ to S ₉ are the same as in FIG.

It is desirable to use FFT (Fast Fourier Transform) as the DFT to shorten the calculation time (claim 3). The block configuration of the measuring device used for detecting the jitter frequency component of the present invention is given in FIG. However, the periodic data fetched in the memory 4a or 4b is the continuous data T ₁ , T ₂ ,
In addition to the case of T ₃ , ..., Sampling data acquired at appropriate intervals as shown in FIG. 4 may be used. In that case, the discrete data P ₁ , P ₂ , ... Obtained by the measurement amount calculation unit 5 are also sampling data (claim 4). Depending on the case, the measured quantity calculation unit 5 may sample some of the periodic data continuously captured in the memory 4a or 4b to calculate the designated physical quantity. Even in that case, the data Pi becomes sampling data (claim 4).

Further, the period measuring circuit 3a or 3b is set for a predetermined time.
Obtain the moving average Tai of the continuous cycle data in T (electronic time
There is a method by the road and a method by the digital operation.
May be provided to the memory 4a or 4b. Measured amount calculation
In part 5, the physical quantity P is calculated from the moving average Tai of the cycle.₁, P₂、…
And ask for P₁, P₂,,… are regarded as approximate moving averages
None, this may be discrete data (claim 5). Place
Depending on the situation, the measured amount calculation unit 5 may move to the memory 4a or 4b.
A predetermined time T
The moving average value Tai is calculated and the physical quantity P is calculated from it.₁，
P₂,,,, P, ₁, P₂,… Is an approximate moving flat
It may be regarded as an average value (Claim 5).

When the frequency spectrum of the jitter is displayed as an image, the peak waveform determined to be aliasing may be deleted and the spectrum X ₁ or X ₂ may be displayed (claim 6). Alternatively, the frequency spectrum X ₁ or X ₂
At the same time, it is possible to display an image of which part of the spectrum is the peak waveform due to aliasing.

The same signal is input to the input terminals INa and INb, the measurement signal generation circuit 2 inverts the polarity of one signal, and the periods T ₁ , T ₂ , ... Of the respective signals shown in FIG. 6A.
Is measured and stored in memories 4a and 4b, and the number of data is 2
It can be doubled. In the above description, the time series digital data {Pi} input to the frequency component analysis unit 6
Although the non-isochronous data is used, it is obvious that the data may be isochronous data.

[0027]

As described above, according to the present invention,
Sampling frequencies fs ₁ and fs ₂ that are slightly different for the time signal x (t) or the time-series data {Pi} are set, data is interpolated and sampled as necessary, DFT calculation is performed, and jitter dispersion is performed. Frequency spectrum X ₁ (nΔ
By comparing the peak frequencies of f ₁ ), X ₂ (nΔf ₂ ), it is possible to easily determine whether or not the peak waveform is due to aliasing.

[Brief description of drawings]

FIG. 1 is a flowchart showing a method for detecting a jitter frequency component of the present invention.

FIG. 2 is a waveform diagram for explaining an example of the present invention.

FIG. 3 is a waveform chart for explaining another example of the present invention.

4A is a waveform diagram of an input signal v (t) of the period measuring circuit 3a or 3b of FIG. 5, and B is a diagram showing data output from those circuits and timing of data acquisition.

FIG. 5 is a block diagram of a jitter measuring device used in the present invention and a conventional jitter frequency component detecting method.

FIG. 6 is a flowchart showing a modified example of the present invention.

FIG. 7 is a waveform diagram illustrating a conventional jitter frequency component detection method.

Claims (7)

[Claims] 1. A continuous time signal x (t) (t is time) is obtained by connecting adjacent data with interpolation curves for time-sequential non-equidistant or equidistant digital data relating to an arbitrary physical quantity. , The first and second slightly different with respect to the time signal x (t)
Sampling frequencies fs ₁ and fs ₂ = fs ₁ ± Δfs are set, and sampling is performed from the time signal x (t) at the first and second sampling periods ts ₁ = 1 / fs ₁ and ts ₂ = 1 / fs ₂ . do it,
First and second discrete time signals (equal-interval discrete data sequence) x
(Nts ₁ ), x (nts ₂ ) (n = 0, 1, 2, ... N-
1) and extract their first and second discrete time signals x (nts ₁ ), x (n
A discrete Fourier transform (DFT) operation is performed on ts ₂ ) to obtain first and second discrete frequency spectra X ₁ (f = 0 to 0).
_{fs 1/2), X 2} (f = 0~fs 2/2) asking them first, frequency f1 ₁ second amplitude of the frequency spectrum X _1, X ₂ becomes a predetermined value or more peaks <f1 ₂ <... <f
1m and f2 ₁ <f2 ₂ <... <f2m are obtained, and the frequencies f1i and f2i (i = 1 to m) of the peaks of the first and second frequency spectra X ₁ and X ₂ are compared to obtain f1i ≈f2i ( A peak having i = 1, 2, ..., M) is determined to be a true peak, and a peak having f2i ≈f1i ± Δfs (i = 1, 2, ..., M) is determined to be a peak due to aliasing. Yes, the method for detecting the jitter frequency component. 2. First and second sampling frequencies fs ₁ and fs which are slightly different from the time-sequential non-equidistant or equidistant digital data {Pi} relating to an arbitrary physical quantity.
₂ = fs ₁ ± Δfs is set, and the first and second data are obtained from the time-series digital data {Pi}.
Sampling periods ts ₁ = 1 / fs ₁ and ts ₂ = 1 /
First and second time series data {Ui}, {Vi} existing in the vicinity of each sampling time of fs ₂ are extracted, and the first and second time series data {Ui}, {Vi} are interpolated. Then, the first and second discrete time signals (equal-interval discrete data strings) x ₁ (nts ₁ ), x ₂ (n have time intervals equal to the first and _second sampling periods ts ₁ and ts ₂ respectively.
ts ₂ ) (n = 0, 1, 2, ... N−1) and extracts their first and second discrete time signals x ₁ (nts ₁ ), x
₂ (nts ₂ ) is subjected to a discrete Fourier transform (DFT) operation to obtain the first and second discrete frequency spectra X ₁ (f
_{= 0~fs 1/2), X} 2 obtains the _{(f = 0~fs 2/2)} , their first, frequency second amplitude of the frequency spectrum X _1, X ₂ becomes a predetermined value or more peaks f1 ₁ <f1 ₂ <... <f1
m and f2 ₁ <f2 ₂ <... <f2m are obtained, and the frequencies f1i and f2i (i = _{1 to} m) of the peaks of the first and second frequency spectra X ₁ and X ₂ are compared to obtain f1i ≈f2i ( A peak having i = 1, 2, ..., M) is determined to be a true peak, and a peak having f2i ≈f1i ± Δfs (i = 1, 2, ..., M) is determined to be a peak due to aliasing. Yes, the method for detecting the jitter frequency component. 3. The method of detecting a jitter frequency component according to claim 1, wherein the discrete Fourier transform (DFT) is performed.
Is a fast Fourier transform (FFT). 4. The method of detecting a jitter frequency component according to claim 1, wherein the digital data is sampling data. 5. The method for detecting a jitter frequency component according to claim 1, wherein the digital data is a moving average value of the physical quantity. 6. The method for detecting a jitter frequency component according to claim 1, wherein the peak characteristic determined to be the aliasing is deleted and the frequency spectrum X ₁ or X ₂ is displayed as an image. Is characterized by. 7. The method for detecting a jitter frequency component according to claim 1, wherein the frequency spectrum X ₁ or X ₂ and which part of the spectrum is a peak waveform due to aliasing are displayed as an image. It is characterized by doing.